Efficient RDO using sample recycling

Deterministic optimization does not consider uncertainties. This may lead to designs which are not robust or reliable. The use of safety factors is the common approach to cope with this problem. The main weaknesses of the achieved results are overdesign (too expensive) or underdesign (unreliable) because safety factors do not necessarily consider the special problem. Therefore robust design optimization uses stochastic values as constraint and/or objective to obtain a robust and reliable optimal design. In classical approaches the effort required for stochastic analysis multiplies with the complexity of the optimization algorithm. The suggested approach shows that it is possible to reduce this effort enormously by recycling previously obtained data. In a simple example, it will be shown that this is possible without loss of accuracy.